smallevals 0.1.0

Small Language Models Evaluation Suite for RAG Systems

smallevals - Small Language Models Evaluation Suite for RAG Systems

A lightweight evaluation framework powered by tiny ( really tiny ) 0.6B models — runs 100% locally on CPU/GPU/MPS, extremely fast and cheap.

Evaluation tools requiring LLM-as-a-judge, that costs/doesn't scale easily. evaluates in seconds in GPU, in minutes in any CPU !

Evaluate Retrieval

Evaluation of RAG system includes retrieval and RAG stage, attacks to test retrieval and RAG answers(in the near future)!

Models

Model Name	Task	Status	Link
QAG-0.6B	Generate golden Q/A from chunks (synthetic evaluation data)	Available	🤗
CRC-0.6B	Context relevance classifier (question ↔ retrieved chunk)	Incoming	—
GJ-0.6B	Groundedness / faithfulness judge (answer ↔ context)	Incoming	—
ASM-0.6B	Answer correctness / semantic similarity	Incoming	—

Current Focus: Retrieval evaluation (QAG-0.5B). Generation evaluation models (CRC-0.5B, GJ-0.5B, ASM-0.5B) are future work.

Installation

pip install smallevals

Quick Start

Evaluate Retrieval Quality (Python)

Connect to your favourite Vector DB (Milvus, Elastic, PGVector, Chroma, Pinecone, FAISS, Weawiate), attach your favourite embeddings, generate questions, and visualise results!

Under the hood, generates question per chunk, and tries to retrieve it as a single-first relevant docs, calculate scores.

from smallevals import evaluate_retrievals, SmallEvalsVDBConnection

vdb = SmallEvalsVDBConnection(
    connection=chroma_client,
    collection="my_collection",
    embedding=embedding
)

# Run evaluation
result = evaluate_retrievals(connection=vdb, top_k=10, n_chunks=200) # Generate question for 200 chunks, and test to retrieve them!

And evaluate results!

Generate QA from Documents (CLI)

smallevals --docs-dir ./documents --num-questions 100

### QAG-0.6B

The model was trained on TriviaQA, SQuAD 2.0, Hand-curated synthetic data generated using Qwen-70B , generating a question from the chunk/doc.

Given the passage below, extract ONE question/answer pair grounded strictly in a single atomic fact.

PASSAGE:
"Eiffel tower is built at 1989"

Return ONLY a JSON object.

{
  "question": "When was the Eiffel Tower completed?",
  "answer": "1889"
}

Known issues:

• Model is trained on text/wiki data, bias towards well structured text.
• Dataset contains question that ask generic questions, dataset will be more carefully crafted in v3.

### Other Models:

Other models to be trained to eliminate the need of external LLMs.

CRC-0.6B : Context relevance classifier (question ↔ retrieved chunk) GJ-0.6B : Groundedness / faithfulness judge (answer ↔ context)
ASM-0.6B | Answer correctness / semantic similarity